Bulk YouTube transcripts: whole channels and playlists

Written by the VidWords Team · June 11, 2026 · Updated July 14, 2026 · Report a correction

Researching a creator, turning a course into notes, training a chatbot on a back catalogue, or auditing competitors — here's how to turn an entire channel or playlist into text.

The 30-second version

Open Bulk extract and pick the Channel or Playlist tab.
Paste an @handle, a channel URL, a UC… id, or any link containing list=…, then hit Find videos — listing is free and covers up to 500 videos.
Hit Extract all. Progress runs three videos at a time, and per-video failures are itemised instead of sinking the batch — they are never charged.
Download the whole run as one .txt, .json, or .csv file.
Automating it? POST /api/channels (Starter and up) turns a handle into video ids; feed those to POST /api/transcripts in batches of 50.

Channel or playlist — which do you want?

Both run through the same bulk extractor and cost the same, so the only question is how the videos are grouped:

A channel gives you a creator's most recent uploads — the right choice for competitor research, or for building a knowledge base from everything one person has published.
A playlist gives you a curated, ordered set — the right choice for a lecture course, a conference track, or a podcast archive, where sequence matters and you don't want the creator's unrelated uploads.

A video that belongs to a playlist can be reached either way; pasting a URL that contains list=… resolves the playlist rather than the single video.

Extracting a channel in the browser

Open Bulk extract and pick the Channel tab.
Paste the channel's @handle (e.g. @mkbhd), its URL, or a UC… channel ID — VidWords resolves the channel ID for you.
Hit Find videos. The latest videos (up to 500) are listed for free — finding videos never costs credits.
Click Extract all. Progress runs three videos at a time; failures (e.g. caption-less videos) are listed per video and never charged.
Download everything as one combined .txt, .json, or .csv.

Extracting a playlist in the browser

Copy the playlist link — anything containing list=… works, including a video URL that is part of a playlist.
Open Bulk extract → Playlist tab → paste → Find videos (free).
Review the resolved list — title, channel, and count, with a note showing exactly how many credits extraction will use.
Hit Extract all, then download the combined file.

What the combined download looks like

Format	Structure	Best for
`.txt`	Each video separated by a heading with its title and link	Reading end to end, or pasting into an LLM
`.json`	Per-video segments with timings and metadata	Scripts, pipelines, and RAG indexes
`.csv`	One row per caption line with video id, start, and duration	Spreadsheets and timestamp filtering

What people build with this

Course notes — turn a 40-lecture playlist into one document you can search across.
Podcast quote mining — search every episode for a phrase, then jump to the exact timestamp.
Competitor and creator research — read what a channel actually says, at volume, instead of skimming titles.
LLM knowledge bases — the combined JSON drops straight into a retrieval pipeline.

Over the API

The channels endpoint (Starter plan and higher) turns handles into video lists, which you then feed to the transcripts endpoint in batches of 50:

curl -X POST https://vidwords.com/api/channels \
  -H "Authorization: Basic YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"ids": ["@mkbhd"]}'

The response includes channelId, the channel title, and each video's id — pipe those ids into POST /api/transcripts.

Limits and practical notes

Channels and playlists both resolve up to 500 videos per request, newest first for a channel and in playlist order for a playlist.
One transcript = one credit. Resolving the video list is always free, so you can check the size of a job before committing to it.
Private and members-only videos can't be read, and some videos legitimately have no transcript (captions disabled, music-only, age-restricted). These come back as per-video errors rather than failing the whole batch, and they are not charged.
Past 500 videos the listing is capped rather than paged — over the API the response comes back with truncated: true so you can tell it happened. For a back catalogue larger than that, split the work into narrower playlists.

Try bulk extraction →